Plant Demolition Simulation From 3D Point Clouds
A web system that ingests scanned point clouds and, through segmentation and support-relationship analysis, physically simulates the collapse of a structure. Built up over three iterative generations.
Duration
Contract development (3 iterative generations)
Team
1-2 people
01 // OVERVIEW
Project Overview
A system that ingests 3D point clouds scanned from a plant and, through preprocessing, segmentation, and GLB export, physically simulates the demolition and collapse of the structure in the browser. Combines a point-cloud-processing backend with a Three.js front end, built out over three generations.
02 // KEY FEATURES
Key Features
Point-Cloud-to-Mesh Pipeline
Reads point clouds such as PLY files and, through downsampling, outlier removal, normal estimation, and DBSCAN segmentation, exports them to GLB. Consistently converts between Z-up and Y-up coordinate systems throughout.
Support-Relationship Analysis and Manual Correction
Automatically detects support relationships between segments, which the user can review and edit. Builds a collapse chain and passes it to the simulation.
In-Browser Physics Simulation
Uses Three.js (R3F) and a physics engine to play back, in the browser, the collapse of a structure that has lost its support.
03 // TECHNICAL DEEP-DIVE
Technical Highlights
Solution
Reconciling Scanner Coordinates With Web Rendering
Explicitly converted between the point cloud file's Z-up coordinates and Three.js's Y-up coordinates at each stage - database storage, API responses, and GLB export - preventing coordinate-system mismatches.
04 // IMPACT & RESULTS
Results
Point Cloud to Collapse
End-to-end pipeline
3 Generations
Built through iteration